AI can compress a task. It cannot create hours that were not worked.
That is the clearest billing rule in ABA Formal Opinion 512, the American Bar Association’s first formal opinion focused on generative AI. A lawyer billing by the hour must bill the time actually spent, including legitimate review and verification. The lawyer cannot charge the historical time a task would have taken without AI.
The opinion was issued on July 29, 2024, but the underlying duties are not new. Competence, confidentiality, communication, supervision, candor, and reasonable fees still apply. AI changes how those duties need to be implemented.
This article is an operational reading of the guidance, not legal advice. Firms should confirm the rules and ethics opinions that apply in every jurisdiction where they practice.
The billing rule is straightforward
Suppose research previously took five hours. An approved AI tool produces a first pass, and the lawyer spends two hours checking every authority, correcting the analysis, and tailoring the result to the matter.
Under an hourly agreement, the billable time is the time actually spent on the client matter. It is not five hours because five hours was the old expectation.
The review work is real work. A lawyer can bill reasonable time spent validating citations, checking the record, applying professional judgment, and revising the output. What the lawyer cannot do is add the machine’s time to the human time or recreate hours that the machine removed.
The same principle applies to cost. If a firm wants to pass through a matter-specific AI expense, the basis needs to be explained and permitted. General software used across the practice is more likely to be treated as overhead. Flat subscription cost should not be converted into a made-up per-matter charge.
AI does not make legal judgment free. It does make fictional time easier to detect and harder to defend.
Flat fees do not remove the reasonableness requirement
A fixed fee can align the firm’s economics with AI-driven efficiency. The client buys a defined service and price, while the firm retains some of the benefit if it delivers efficiently.
That does not mean any fixed fee is automatically reasonable. Model Rule 1.5 still applies. Scope, complexity, risk, local rules, the lawyer’s experience, and the value of the service remain relevant. If AI changes the cost and effort of a service materially, firms should review whether the existing price and explanation still make sense.
The durable approach is to define the service, set a clear scope, explain the fee, and document what happens when work falls outside that scope.
The ethics analysis starts before billing
Billing gets attention because it is visible to the client. The more consequential failure can happen earlier, when client information enters a tool without adequate review.
Formal Opinion 512 ties generative AI to several Model Rules.
Competence under Rule 1.1
A lawyer needs a reasonable understanding of the tool’s capabilities and limits. That includes knowing that a model can invent authorities, omit controlling law, lose context, or produce a fluent answer that is wrong.
Competence does not require becoming a machine-learning engineer. It does require enough understanding to choose an appropriate use, design a review process, and recognize when the output cannot be trusted.
Confidentiality under Rule 1.6
Before client information enters an AI service, the firm needs to understand how the vendor handles prompts, files, outputs, logs, retention, access, training, subprocessors, and deletion.
A consumer account and an enterprise account may have different terms even when the product name looks the same. Firms should verify the exact plan, contract, settings, and data flow rather than relying on a vendor’s general marketing page.
Communication under Rule 1.4
Client disclosure depends on the facts, the jurisdiction, the engagement, and how AI affects the representation. A firm should not reduce this to a universal sentence that says disclosure is always required or never required.
The practical solution is an engagement process that prompts the responsible lawyer to decide whether consent or disclosure is needed and records that decision.
Supervision under Rules 5.1 and 5.3
AI output still needs a responsible lawyer. The duty does not disappear because the system appears autonomous or because a vendor calls it an agent.
The reviewer needs a defined task, the authority to reject the output, and a checklist that matches the risk. A signature at the end of the process is not meaningful if the reviewer cannot see what the system did.
Fees under Rule 1.5
The fee must be reasonable. Hourly billing must reflect actual time. Costs must be explained and handled consistently with the engagement and applicable rules.
Florida shows how local guidance can add detail
Florida Bar Ethics Opinion 24-1 addresses confidentiality, competence, fees, and advertising. It specifically warns against double billing and falsely inflated time. It also explains that a lawyer may charge only actual costs incurred for an individual client and should not duplicate overhead.
The opinion is advisory and specific to Florida, but it illustrates why a national policy is not enough. A firm needs a jurisdiction matrix that tracks the rules, opinions, standing orders, and client requirements relevant to its lawyers and matters.
Build the rule into the workflow
An ethics memo stored in a shared drive will not control day-to-day behavior by itself. The rule needs to appear where the decision happens.
At matter intake
Record the governing jurisdictions, client AI restrictions, confidentiality level, approved tools, and required disclosures.
Before an AI tool is used
Require a matter number, user identity, approved use case, and data classification. Block sensitive data from tools that are not approved for it.
During review
Save the source material, output, reviewer, corrections, and approval. The evidence should be proportionate to the risk, but high-risk work should not rely on memory.
At time entry
Record the actual human time. If the client requires AI disclosure in the narrative, make that a structured billing rule instead of expecting every timekeeper to remember it.
At pre-bill review
Check the engagement, outside counsel guidelines, matter approvals, AI-related costs, and time narratives before the invoice leaves the firm.
What an engagement process should cover
Firms often ask for a single AI clause. The clause matters, but the decisions behind it matter more.
Before using standard language, decide:
- which AI uses the firm permits
- whether client information may be processed
- what vendor safeguards apply
- when client consent is required
- who reviews AI-assisted work
- how fees and costs will be calculated
- how a client can restrict or opt out of a use
- how records will be retained
Counsel can then tailor the engagement language to the jurisdiction and matter. The goal is not to promise that AI is risk-free. It is to make the firm’s process accurate, understandable, and enforceable.
What firms can charge for
The answer is more useful when framed positively.
Firms can charge reasonable fees for legal services. Under an hourly agreement, that includes actual time spent applying judgment, verifying authorities, reviewing facts, correcting output, and communicating with the client. Under an agreed fixed or value-based structure, the fee can reflect a defined service rather than elapsed time, subject to the applicable reasonableness rules.
Matter-specific third-party costs may be chargeable when they are actual, explained, permitted by the engagement, and not already included in overhead. The firm should be able to show how the amount was calculated.
What does not work is inventing time, double billing, or applying a technology surcharge with no clear basis.
A practical compliance checklist
- Identify every AI tool used for client work.
- Separate consumer, enterprise, and private deployments.
- Review vendor contracts and data settings.
- Publish an approved-tools and approved-uses list.
- Add client and jurisdiction rules to matter intake.
- Define review standards by use-case risk.
- Record actual time and actual matter-specific cost.
- Check engagement terms and client guidelines before invoicing.
- Review the policy when tools, terms, or guidance change.
AI billing ethics is not a single policy question. It is a system design problem. The legal rule, the vendor contract, the matter workflow, and the invoice need to agree.
Need the policy to work in practice?
Jinka helps firms connect AI governance to identity, data access, matter workflows, review records, and billing controls.